Topology-Aware Load Balancing in Datacenter Networks

Datacenter networks play a vital role in accommodating massive data transmission, such as in Giga/Terabytes on multiple links. With this inevitable data transmission, some uncertainties like dynamic traffic behavior, topology asymmetry, and link failure are major issues that data center faces, resulting in poor performance. The load balancing scheme should be accurate, especially path sensing and timely responses to mitigate the data center network's fallouts. Different solutions have been proposed to overcome these issues. Some of the proposed techniques are unaware of path conditions. They reroute at a fixed data rate blindly. Others could sense the path condition but reroute and adjust the load, balancing very slowly, degrading the performance. A few research studies demand hardware changes, particularly for acquiring accurate congestion feedback from the end host and adopting load balancing. Thus they can not react timely to these uncertainties, which results in reasonable performance degradation, especially in the worst situation to detect some issues like random packets drop and throughput of the channel. In this paper, a proposed solution focuses on two things (1) accurate timely acknowledgment of packet delivery, (2) Current load on each link. In our proposed work, there is no need for hardware changes. In our proposed scheme, the accurate response of acknowledgment of each packet delivery is achieved. The proposed scheme has been implemented using the NS3 simulator. The results of our proposed scheme have been compared with the other existing state of the art solutions.